currents
Noncontact protein arrayer
onto a slide. Laser pulses are coordinated with the moveProtein samples often are available in limited volumes and ments of the target and the slide to create a protein microarray. can become clogged in commercial arrayers that were deThe spot volume can be adjusted by changing the laser enersigned for printing DNA solutions. gy. Because BioLP deposition is a So Bradley Ringeisen and co-worknoncontact process, multiple proLaser absorption layer ers at the Naval Research Laboratein solutions can be printed from Energy meter CCD camera tory, the University of Illinois at Urmultiwelled targets without washSupport bana–Champaign, and Expression ing cycles. layer Laser Pathology, Inc., have developed a To test the function of spotted noncontact protein-arraying deproteins, the researchers used the Microscope objective vice. Called biological laser printBioLP to print a bovine serum alProtein solution ing (BioLP), the new arrayer deposbumin (BSA) microarray. When the Target support its proteins in 30-μm-diam spots array was incubated with a fluoresthat contain ~500 fL of liquid, which cently labeled anti-BSA antibody, Microarray substrate all of the spots became fluorescent. is much less material than a commercial model can print. In another experiment, Ringeisen Look, ma, no hands. The BioLP device prints protein spots BioLP transfers liquid by a fluand co-workers printed an alkaby a noncontact fluid jetting mechanism. (Adapted with id jetting mechanism. The protein line phosphatase (AP) array onto permission. Copyright 2005 Wiley-VCH Verlag GmbH & Co.) solution to be printed is absorbed a slide and then printed AP substrates onto the spots. Positive sig onto a target, which is an optically transparent quartz disk coated with metal or metal oxide. nals were observed at every spot. Thus, the researchers say When a pulsed laser hits the back of the target, the liquid on that BioLP-spotted proteins retain their enzymatic and bindthe absorption material directly under the laser pulse drops ing capabilities. (Proteomics 2005, 5, 4138–4144)
Versatile mass spectrometer By combining the front end of a commer cial triple quadrupole (qQq) instrument with a Fourier transform ion cyclotron resonance (FTICR) mass analyzer, Pe ter O’Connor and colleagues at the Bos ton University School of Medicine and MDS Sciex (Canada) created an electro spray qQq FTICR mass spectrometer for biological applications. The new instru ment was designed with proteomics in mind, in particular for studying post translational modifications and for topdown analysis of protein samples. The front end of the instrument con tains a focusing rf-only quadrupole, a re solving quadrupole, and a LINAC quad rupole collision cell (Q2). When these are combined with an FTICR mass ana lyzer, the system can detect low-abun dance species. In addition, it is compat ible with a wide range of fragmentation techniques, including nozzle-skimmer fragmentation, Q2 collisionally activat ed dissociation, multipole storage-as sisted dissociation, and electron-cap ture d issociation. To demonstrate the performance of the instrument, the researchers used it to analyze phosphorylated peptides. They also performed top-down sequenc
ing on several proteins, including com mercially available and biologically de rived ones. The top-down approach allowed for the unambiguous identifica tion of the human E2 ubiquitin-conju gating enzyme, UbcH10. According to the researchers, one of the key features of the instrument is its versatility for addressing numerous bi ological problems. Other advantages in clude increased selectivity of precursor ion species, sensitivity, and speed. ( J. Am. Soc. Mass Spectrom. 2005, 16, 1985–1999)
A sense of structure Each of the two main methods for deter mining protein structure—NMR spec troscopy and X-ray crystallography—has distinct limitations and benefits. To de termine whether one method is prefer able to the other or whether the two can serve complementary purposes, two groups involved in the Northeast Struc tural Genomics Consortium screened >400 proteins by both methods. The researchers produced 1H–15N het eronuclear single-quantum correlation (HSQC) spectra of each protein, believ ing that well-resolved HSQC spectra in dicated folded proteins that were ame nable to NMR spectroscopy. They then
10 Journal of Proteome Research • Vol. 5, No. 1, 2006
categorized the results as excellent, good, promising, or poor and/or un folded on the basis of parameters such as spectral dispersion, line widths, and number of resolved peaks. Similarly, they screened each protein for its abili ty to produce diffraction-quality crystals and, when possible, they determined 3‑D structures. Of the 420 proteins studied, the re searchers could determine the struc tures of only a handful by both meth ods. Furthermore, although good HSQC spectra did indicate how well a protein was folded, the spectra were poor indi cators of whether the protein could form crystals or whether its structure could be determined by X-ray crystallography. At the same time, although each method was applicable to only a sub set of the proteins studied, the research ers found that a concerted approach with both methods increased the likeli hood of a 3-D structure being calculated for a given protein. Thus, the groups rec ommend that a parallel approach with both NMR spectroscopy and X-ray crys tallography be pursued to maximize the success rate of any large-scale structur al study. ( J. Am. Chem. Soc. 2005, 127, 16,505–16,511; 16,512–16,517) © 2006 American Chemical Society
currents Malaria parasite interactome A parasite called Plasmodium falcipar um kills millions of people each year and is responsible for >90% of human deaths from malaria. Despite the importance of this human pathogen, little is known about its proteins. Several research groups are working to change that. In a pair of recent papers, researchers report the first steps toward understand ing the protein interaction network for P. falciparum. The data reveal previously unknown pathways and processes of the malaria parasite and could lead to new drug and vaccine targets. In the first paper, Robert Hughes, Stan ley Fields, and colleagues at the Howard Hughes Medical Institute, the Universi ty of Washington, and Prolexys Pharma ceuticals used a high-throughput yeast two-hybrid assay to identify 2846 P. fal ciparum protein–protein interactions. Most of the interactions involved at least one previously uncharacterized pro tein. A connectivity analysis of the data revealed interacting protein groups, in cluding one involved in chromatin mod ification, transcription, mRNA stability, and ubiquitination and another involved in the invasion of host cells. In an accompanying paper, Silpa Suthram and colleagues at the Univer sity of California, San Diego, examined how the P. falciparum protein interac tion map compares with those of other
Ovarian cancer biomarkers
species, including the yeast Saccharomy ces cerevisiae, the nematode worm C. el egans, the fruit fly Drosophila melano gaster, and the bacterium Helicobacter pylori. The researchers found that the protein interaction network of P. falci parum has little in common with that of the other organisms. The results are not too surprising, as the genome of P. falci parum contains thousands of encoded proteins that are not found in other eu karyotes. The lack of conservation be tween the P. falciparum protein network and those of other species could eventu ally facilitate the discovery of new drugs targeted at the parasite’s unique protein complexes. (Nature 2005, 438, 103–107; 108–112)
Biomarkers of benzene exposure Inside a shoe factory in Tianjin, China, workers are exposed to benzene levels in the air that often exceed 30 ppm. Could such exposure lead to cancers such as acute myeloid leukemia? To better un derstand the toxicity of benzene and identify biomarkers of exposure, Roel Vermeulen and colleagues at the Na tional Cancer Institute; the University of California, Berkeley; Ciphergen Bio systems; and the Chinese Center for Dis ease Control and Prevention examined the serum proteomes of healthy exposed shoe-factory workers and unexposed
with advanced ovarian cancer and compared Proteins that are difthem with those in ferentially expressed control samples from could serve as bio11 healthy women. The markers for ovarian proteins were separatcancer, according to ed by SDS-PAGE and new results published identified by MALDI by Nuzhat Ahmed and TOFMS and western colleagues at the Royblotting. All proteins al Women’s Hospital, were analyzed after Protein isoforms. 2DE profile of the University of Melthe depletion of highserum from (a) a healthy woman bourne, Baker Heart abundance albumin. and (b) a grade 3 ovarian cancer Research Institute, A series of haptopatient. (Adapted with permis Monash University (all globin and transferrin sion. Copyright 2005 Wiley-VCH in Australia), and the precursors were found Verlag GmbH & Co.) University of Leipzig to be either up-regu(Germany). The biolated or down-regumarkers could lead to earlier detection lated in the ovarian cancer patients. of the disease and provide a means for The expression of these proteins varmonitoring drug treatment. ied depending on the grade of canThe researchers analyzed proteins cer and changed after chemotherapy. in 24 serum samples from women (Proteomics 2005, 5, 4625–4636)
Toolbox N-linked glycopeptide analysis To study N-linked glycosylation, researchers typically deplete samples of peptides that are not glycosylated. The remaining glycopeptides are then treated with peptide N-glycosidase F, which removes the carbohydrates from the peptide backbone; this process also results in the deamidation of the asparagines. False positives are often obtained, however, when scientists mine available databases by allowing for asparagine deamidation in their searches, so all the spectra must be manually validated. To eliminate the need for manual validation, James Atwood III and colleagues at the University of Georgia and the Windber Research Institute have developed and tested an algorithm that modifies database entries so that only proteins that contain an N-linked glycosylation consensus sequence are searched. When proteins that do not contain the consensus sequence are filtered out, the effective database size is also decreased and the likelihood of obtaining false positives is reduced. The algorithm can be downloaded at http://128.192.9.86/stargate/ ModifyDB.zip. (Rapid Commun. Mass Spectrom. 2005, 19, 3002–3006)
Signal maps Benno Schwikowski and colleagues at the University of Washington, the Institute for Systems Biology, the Fred Hutchinson Cancer Research Center, the Swiss Federal Institute of Technology, the University of Zurich, Wesleyan University, Institut Pasteur (France), and Cedars-Sinai Medical Center have developed an approach to compare the protein compositions of different biolog ical samples. In the new approach, the researchers use an algorithm to construct signal maps that align LC data on the basis of raw MS signals obtained from multiple experiments. A featuredetection algorithm then identifies signals that are similar or different among the experiments. The researchers demonstrated that biomarkers can be discovered with this method by analyzing two standard samples with slightly different compositions. (Mol. Cell. Proteomics 2005, doi 10.1074/mcp. M500133-MCP200)
Journal of Proteome Research • Vol. 5, No. 1, 2006 11
currents Toolbox More confident protein IDs By combining the recently published average peptide score (APS) method with prefiltering of peptide identifications, Ian Shadforth and colleagues at Cranfield University, Silsoe; Cambridge University; and GlaxoSmithKline (all in the U.K.) have reduced the number of false-positive protein identifications. In the new dual-threshold approach, peptide identifications with low Mascot scores are filtered out. The APS threshold is then set on the basis of the maximum APS returned from a reversed database search. The filtering and APS thresholds are iteratively adjusted to determine the levels at which false positives are reduced to almost zero. The method was demonstrated on Arabidopsis thaliana membrane protein samples. The researchers say the method works well but is time consuming because each search is run two times: once against a forward database and once against a reversed database. Alternatively, the APS filter can be calculated on the basis of the Mascot “plughole” score. The use of the plug hole score as a filter saves time but is not as effective. (Rapid Commun. Mass Spectrom. 2005, 19, 3363–3368)
Preprocessing LC/MS data In liquid chromatography (LC)/MSbased proteomics experiments in which multiple samples are analyzed in parallel, it is important to recognize the peak variation of each molecule from one sample to the next. Xiang Zhang and colleagues at Purdue University and Beth Israel Deaconess Medical Center have developed software to do just that. Called Xalign, the new software uses a series of algorithms for preprocessing LC/MS data. A spectral deconvolution algorithm processes LC/MS data down to the peptide ion level. The program improves the analysis of peptide samples by using chemical noise filtering, charge-state fitting, and de-isotoping. A two-step peak alignment process adjusts for retention time drift between samples and aligns all samples to the median sample. A combination of tests, including a Kolmogorov–Smirnov test, a peak-number test, and an alignmentquality test, is performed at several different steps for data quality assurance. (Bioinformatics 2005, 21, 4054–4059)
controls selected from a nearby clothing manufacturing plant. Serum samples from 10 exposed work ers and 10 controls were fractionated by anion exchange and bound to 3 types of protein arrays—hydrophobic, metal af finity, and cation exchange. Protein ex pression patterns were then obtained by SELDI TOFMS. The researchers found three proteins (4.1, 7.7, and 9.3 kDa) consistently downregulated in the exposed workers com pared with the controls. All proteins were inversely correlated with individu al estimates of benzene exposure, which were determined by the subjects wear ing organic vapor passive monitoring badges. The 7.7- and 9.3-kDa proteins were subsequently identified as platelet factor 4 and connective tissue activating peptide-III by MS/MS analysis of proteo lytic fragments. Although the study involves only a limited number of samples, the results suggest that the down-regulation of CXC chemokines may play a role in benzene toxicity and could serve as a biomarker for early detection of benzene exposure. (Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 17,041–17,046)
Accelerating proteomics analyses High-performance liquid chromatogra phy (HPLC)/MS and HPLC/MS/MS have become preferred methods for largescale proteomics experiments. Inher ent limitations in throughput, however, present roadblocks to scientists looking to analyze large numbers and varieties of samples. Two independent research groups recently tackled this throughput problem. Yingming Zhao and colleagues at the University of Texas Southwestern Med ical Center at Dallas examined the im pact of different buffers and gradient profiles in nanoflow HPLC on peptide identification by MS and MS/MS. By loading protein digests in low-organic buffer but running the column and per forming an isocratic gradient in high-or ganic buffer, they could generate peak widths of 4–6 s, more than sufficient for high-speed MS and MS/MS analyses. Furthermore, the method reduced anal ysis times to ~5 min, allowing the scien tists to generate data from 96 in-gel di gests in one 8-h working day. Richard D. Smith and colleagues at the Pacific Northwest National Labora
12 Journal of Proteome Research • Vol. 5, No. 1, 2006
tory and at the Johannes Gutenberg-Uni versität Mainz and Merck KGaA (both in Germany) enhanced mass spectral throughput by replacing the tradition al stepwise MS/MS analysis with the ac curate mass and time (AMT) tag meth od, which combines HPLC elution times with accurate mass measurements to identify peptides from an AMT tag data base. The switch to AMT allowed the re searchers to obtain good results from both Fourier transform ion cyclotron resonance (FTICR) MS and TOFMS. Using high-speed capillary HPLC combined with high-accuracy MS mea surements, the researchers identified ~2000 different peptides and ~600 dif ferent proteins from a tryptic digest of a Shewanella oneidensis extract within 2– 3 min. They also determined that FTICR MS worked better for analysis times >150 s, whereas TOFMS was most efficient for times 5 kDa) occurs in the first TOF analyz er. This step is then followed by reaccel eration and high-resolution mass anal ysis of the fragments in a reflectron TOF analyzer. The researchers designed special soft ware to compare experimental MS/MS spectra with in silico-generated MS/MS spectra from all protein sequences in a proteomics database. Microorganisms were identified by inference on the basis of individual protein biomarkers. Using this approach, the researchers unambiguously identified intact B. atro phaeus and B. cereus spores in both pure form and in mixtures. The researchers plan to extend the approach for rapid vi rus or protein toxin identification. (Anal. Chem. 2005, 77, 7455–7461)
